What research ecosystems have to learn after everyone agrees to collaborate
On August 21, 2026 I had the privilege of participating in the launch of the Amity Global Research Hub at the Consulate General of India in New York and of joining AGRH’s Advisory Committee. The gathering brought together scientists, academic leaders, healthcare executives, technologists, industry representatives, and institutional leaders from India and the United States. Much of the conversation centered, appropriately, on collaboration: across institutions, countries, disciplines, sectors, and scientific traditions.
The event left me thinking less about whether collaboration is important than about what we mean when we claim to have achieved it.
Most institutions already understand that a memorandum of understanding, a conference, or a new center does not by itself constitute meaningful collaboration. The problem is subtler. Those activities are visible, countable, and relatively easy to celebrate, while the slower work of integrating expertise, reconciling incentives, establishing governance, generating evidence, and translating results into use is much harder to see. We can understand the distinction intellectually and still build measurement systems that reward the former more readily than the latter.
That tension is not a criticism of AGRH, which is only beginning its work. In fact, what made the AGRH launch interesting was that many speakers were already wrestling with exactly these questions: how to move beyond relationship formation toward translational work and accountable outcomes. The event prompted the broader question for me because it applies to any institution ambitious enough to describe itself as a research and innovation hub.
My shorthand for the question is simple:
A hub is a verb.
The point is not that places, institutions, affiliations, and networks are unimportant. They create the conditions under which collaboration can occur. The harder test is whether the ecosystem develops an operating capability: whether it can repeatedly convert relationships into productive teams, scientific capability into reliable knowledge, and, where translation is the objective, promising ideas into outcomes that work under real-world conditions.
The word repeatedly matters. What should become repeatable is not the scientific result. Research is not manufacturing, and the next important problem may look nothing like the previous one. What can become repeatable is the discipline through which people form teams, interrogate assumptions, generate evidence, make decisions, translate where appropriate, learn from failure, and carry that learning into the next problem.
Connection is an input. Collaboration is a capability.
Institutional partnerships begin with relationships, and relationships matter. A conversation can uncover complementary expertise. An MOU can create permission to work together. A conference can expose participants to people, methods, and questions they would not otherwise encounter. Weak ties and serendipitous encounters are often sources of genuinely novel ideas.
The mistake is not creating these connections. It is treating their existence as sufficient evidence that an ecosystem can do consequential work.
That capability appears when a sufficiently important problem begins to organize the relationship. Participants have to develop some shared understanding of what they are trying to learn or accomplish, what each party contributes, what evidence will be persuasive, where decisions reside, and how the group will respond when emerging evidence undermines the initial plan. Collaboration becomes more demanding when disagreement is no longer a threat to the partnership but part of how the partnership learns.
This is also where the distinction between multidisciplinary and transdisciplinary work becomes useful. Multiple disciplines can contribute to a project without materially changing one another’s understanding of the problem, and sometimes that is exactly what is needed. Other problems resist disciplinary partitioning. In those cases, academic expertise may need to interact not only across fields but with the knowledge of practitioners, patients, users, policymakers, industry participants, communities, and others who understand the system from different positions within it. Their contribution can change the question being asked, not merely validate a solution after the fact.
Healthcare AI makes the point visible. A predictive model may begin as a problem in computer science, statistics, and data engineering. Whether it improves care can depend on clinical workflow, patient behavior, regulation, reimbursement, data provenance, ethics, implementation, organizational incentives, and whether a clinician operating under time pressure trusts what appears on the screen. None of those considerations makes the model less scientific. They reveal that the object being optimized and the system in which it must create value are not the same thing.
For problems like these, the first product of the ecosystem may not be technology at all. It may be trust organized around a consequential problem.
Different ambitions require different standards of evidence
There is an important risk in arguing too strongly for translation and impact: not all research should be judged by whether it produces an immediately adoptable solution.
Fundamental research creates knowledge whose eventual applications may be unknowable at the time of discovery. A new method, dataset, theoretical insight, negative result, trained scientist, or better scientific question can be an entirely legitimate outcome. Forcing every research program to articulate a near-term commercialization or implementation pathway would distort science rather than strengthen it.
A research hub and an innovation ecosystem therefore make related but different claims. A research enterprise may reasonably say, “Our objective is to generate new knowledge.” An explicitly translational initiative is making an additional promise: that at least some of that knowledge is intended to move toward use. A program promising societal impact is making a stronger claim still.
The operating discipline should match the claim.
If the objective is discovery, evaluate the quality and significance of the discovery. If the objective is translation, then the conditions required for translation belong inside the design. If the claim is improved health, economic value, environmental benefit, or some other societal outcome, publications and pilots may be important milestones, but they cannot substitute indefinitely for evidence that the claimed outcome occurred.
At the same time, not every valuable outcome is immediate, quantifiable, or attributable to one institution. Research may create value years later, influence work elsewhere, or contribute to an outcome that no single participant can reasonably claim. The discipline is not to force false precision, but to be explicit about the kind of value being pursued and the evidence that would reasonably support that claim.
“Impact” must not become an all-purpose word that means everything and therefore measures nothing.
Translation is a feedback loop, not the end of a pipeline
My experience at Northwell Health made the translation problem particularly visible. I worked across clinical data, AI, investment, commercialization, venture creation, and implementation. The closer an idea moved toward real use, the more apparent it became that technical performance was only one component of value.
A model could perform well and still fail because it entered the wrong workflow. A prototype could attract enthusiastic sponsors and then disappear because no one owned the operational transition. A scientifically credible solution could stall because governance, incentives, data availability, economics, or human behavior had been treated as downstream implementation details rather than part of the original problem.
The easy response is to say that innovators should “think about implementation earlier.” That is true, but it still suggests a one-directional pipeline: research first, implementation later, with earlier consideration of the latter.
The reality is more recursive.
Implementation can expose a scientific weakness. User behavior can reveal that the original outcome was poorly specified. A governance constraint can change what data are realistically available. Real-world performance can force a model back into development. A failed pilot can generate a better research question than the one that produced the pilot.
Translation, in other words, sends information upstream.
That is why a useful translational ecosystem is not simply good at pushing ideas from discovery toward practice. It is good at moving learning in both directions. Evidence reshapes the collaboration. Practice reshapes the scientific question. The system learns not only whether an artifact works, but what the environment is telling us about the artifact.
The value of a framework is therefore not that every initiative should move neatly through stages. It is that a serious ecosystem develops the ability to move deliberately amongst stages rather than confusing motion in one direction with progress.
Evidence gives partnerships permission to revise themselves
Large collaborations have another predictable difficulty. Once respected institutions, senior leaders, resources, and reputations become attached to an initiative, revising the underlying hypothesis can become socially and organizationally expensive.
At SIYOM, I use Constructive Inquiry to describe a discipline intended to counter that tendency: test rather than tell, prefer evidence to assertion, and treat iteration as the mechanism through which we get righter, faster.
Too often forced certainty is the requisite for action. Demanding complete evidence before beginning would create paralysis, especially in genuinely novel work where the necessary evidence can exist only after experimentation. The more useful purpose of evidence is to create a legitimate mechanism for revision.
The point is not a generalized skepticism; a partnership should be able to distinguish what it knows from what it assumes. It should know what observations would strengthen its confidence and what observations would weaken it. It should be able to discover that an attractive technology is attached to the wrong problem, that a proposed user does not actually need the solution, or that an implementation barrier changes the economics enough to justify a different approach.
This becomes particularly important once collaborative momentum develops. Institutional enthusiasm can create its own form of Certainty Theater: after enough people and resources have been committed, confidence becomes easier to reward than reconsideration.
Constructive Inquiry is useful precisely because it treats changing direction in response to evidence as progress rather than embarrassment. The objective is not to eliminate uncertainty. It is to build an ecosystem capable of learning faster than its assumptions harden.
Go deeper before you go bigger
New research networks naturally benefit from breadth. More institutions, disciplines, and relationships create more combinations, more weak ties, and more chances for an unexpected opportunity to emerge. During exploration, that optionality is valuable.
The tension appears when a network moves from exploration toward execution. Breadth then carries coordination costs: additional governance, incentives, data questions, intellectual-property issues, and ambiguity about ownership. A portfolio can expand much faster than the system’s capacity to manage it.
That is where go deeper before you go bigger becomes useful: not as a universal rule, but as a discipline for execution. A small number of consequential problems can teach a new ecosystem how to assemble the right teams, establish useful evidence standards, resolve governance questions, translate results where appropriate, and understand why the work did or did not produce value. Expansion then becomes the application of a learned operating discipline rather than the multiplication of projects.
That framing may be particularly useful in India-U.S. collaboration. The strongest partnerships will do more than exchange complementary assets. They will allow both sides to influence the problem definition, research design, evidence standards, and pathway to use. Co-development becomes meaningful when participation changes the work itself.
What should become repeatable is the capacity to learn, adapt, and deliver
A consequential research ecosystem does not need every project to become a product, every relationship to become a transdisciplinary program, or every discovery to demonstrate near-term social impact. Those would be artificial standards, and applying them indiscriminately could suppress exactly the exploratory science and serendipitous relationships a research hub should enable.
The harder discipline is matching the operating model to the promise being made.
If the promise is exploration, create conditions for exploration and judge it accordingly. If the promise is discovery, value rigorous discovery. If the promise is translation, design translation into the work. If the promise is societal impact, define what that means well enough that activity metrics cannot quietly take its place.
For initiatives that do aim to move from scientific possibility toward practice, three principles remain useful:
- Choose consequential problems rather than beginning with fashionable technologies;
- Introduce evidence, governance, implementation, adoption, and stakeholder knowledge early enough to shape the work; and
- Evaluate not only what happened, but what the ecosystem learned that improves its ability to address the next problem.
A successful project tells us that something worked once. A mature ecosystem can explain why it worked, what almost prevented it from working, which assumptions proved wrong, what changed during translation, and how that learning improves its ability to produce appropriate results across different problems.
That is what should become repeatable: not the result itself, but the capacity to learn, adapt, and deliver.
And that is what I mean when I say a hub is a verb.
The noun gives us the institution, the network, the convening power, and the possibility. The verb is what happens when those assets begin to work together: connecting where connection is useful, integrating where the problem demands it, testing assumptions, generating evidence, translating in both directions, learning, adapting, delivering, and occasionally deciding that an idea should stop.
The names around the table matter. The science matters. The relationships matter.
What the ecosystem learns to do with them matters more.
-Marc d. Paradis
About the Author: Marc d. Paradis’ professional journey is a fusion of academic rigor with real-world impact. He began his career over 30 years ago as an academic molecular neurobiologist, instilling in him a deep respect for critical thinking and the scientific method.
Transitioning into industry, he held leadership roles that bridged data and healthcare: as Vice President of Data Strategy at Northwell Health, Marc leveraged one of the world’s most diverse clinical data sets to drive patient-centered innovation via a $100M partnership with Aegis Ventures, launching multiple AI-centered startups; and as Vice President & Dean of Data Science University at Optum, he spearheaded the training of thousands of professionals in practical, product-centric AI, data-driven decision making, and ethical data practices. In each role, he fostered cultures of curiosity, critical thinking, and collaboration – precursors to the Constructive Inquiry ethos.
About SIYOM Consulting: Founded by Marc d. Paradis, SIYOM Consulting is a boutique advisory specializing in Data and AI Strategy for Healthcare and Life Sciences. We help health-system executives, pharma innovators and investors identify, evaluate and execute on high-value data and AI opportunities.